Review: Server error: RangeError: Out of memory

[digitalmaster]

Running plannotator review results in:

❯ plannotator --version
plannotator 0.20.3
❯ plannotator review
[plannotator] Server error: RangeError: Out of memory

[digitalmaster]

This fixed it for me locally. Here's rationale agent shared:

Updated ~/.plannotator/config.json:
"defaultDiffType": "unstaged"
Why this fixes it: upstream Plannotator’s all mode does git diff <empty-tree>..HEAD, which renders every tracked file as an added file. In this repo that is massive, so it can OOM.
Verified :
- plannotator review reached ready state at http://localhost:54416 with no error.
- plannotator opencode-review also reached ready state at http://localhost:54459 with no error.
I stopped both after readiness because review commands normally wait for browser feedback.

[cschanhniem]

A RangeError: Out of memory in Node.js typically means the V8 heap limit was hit (default ~2 GB for 64-bit). A few diagnostic questions that would help narrow the cause:

1. How large is the diff/repo you are reviewing? (lines changed, file count)
2. Does the error happen consistently on any repo, or just a specific one?
3. What is the output of plannotator --version? (already provided: 0.20.3)

Potential causes to investigate:

• git diff output could be very large and the JSON serialization of the parsed diff might exceed V8 heap limits
• If plannotator loads all diff data into memory before rendering, a streaming/batched approach could help
• A memory leak in the websocket or file-watching layer could accumulate over time

A quick workaround is to increase the Node.js memory limit:

NODE_OPTIONS="--max-old-space-size=4096" plannotator review

If that resolves it, the issue is heap size — and the fix would involve either streaming the diff data or adding a max-file-size guard with a user-facing warning.

[digitalmaster]

How large is the diff/repo you are reviewing? (lines changed, file count)

Diff was small (+1,151 −0). But repo is large (31,759 tracked files).

[cschanhniem]

Thanks for the datapoint, @digitalmaster. 31k tracked files with a small diff confirms this is an OOM caused by git scanning the full repo history in all mode, not by diff size. Using "defaultDiffType": "unstaged" is a solid workaround. The real fix would be for Plannotator to cap the scope of all mode or stream the diff instead of loading it entirely into memory. Worth filing a feature request upstream?